Reputation
Badges 1
11 × Eureka!Hi John, but the task is running the script code, so how can that be?
OK so I hope this is not a too vague question, but can you tell me, what may be the benefits of still using clearML in this mode?
@<1523701087100473344:profile|SuccessfulKoala55> Yes, kind of, except that they do use SLURM. I am really trying to understand what are the benefits of using open source clearML in this case
Hi John. It turns out that the task that ended was the pipeline step. Apparently since I create the task that I then run with clearml-agent execute, in the step, it has to remain running. I added task.wait_for_status() at the end of the step and now my pipeline works.
I have another question: Do you know if it is possible to use PipelineDecorator.component as a class decorator?
Hi John.
My E-mail address is oren.shani@mail.huji.ac.il . I tried again and I get the same problem (see attached screenshot).
Oren
OK sorry about that... Let me explain. We have researchers that sometimes work alone and sometimes in teams and our compute resources are in a SLURM based cluster. So basically, with the open source version all they can do is to open projects and run tasks while relaying on the system services to enable access separation, etc. So how is this better than just keep using "normal" python?
@<1523701070390366208:profile|CostlyOstrich36> , After talking to my manager about this, can someone from marketing contact me about Enterprise ClearML pricing?
Thank you John, I suspected so...
So when using the open source clearML server, what are my options as far as multi-tenancy is concerned? I mean if I users simply run cleaml-init, with the credentials of a specifc project, and then clearml-agent with queue names, etc, How do I enable user/groups separation?
Hi John, okay I found it now. Thanks
OK, thank you. I think the only real limitation is that the GUI will probably not be useful for our users, but they can do everything (create projects, tasks, queues etc.) from the command line, right?